can you suggest when to stop training?

[BarCodeReader]

Hi,

Thanks for your code!

Recently I need to re-train this network, on another dataset, HKU-IS, the model converges good,
but I don't know when to stop it...

the loss now in your code is 0.099 and 2.83 respecitively...
Because we don't have evaluation during the training...thus I do not know how good is the result now...
can you tell us when to stop the training when you train the BASNet using DUT-TR dataset?

[xuebinqin]

Thanks for your interests. This is a good question. What we tried before is plotting the -log(loss) of the training process. If the training curve is flat, it should be ok to stop. In your case, if you just want to get better results, you can also pick one of the other SOD datasets, e.g. MSRA, ECSSD and so on, as your validation set. Then select the best model based on the validation results. We also gave our suggested iteration numbers (around 400K on DUTS-TR dataset) in the paper.

…

On Tue, Sep 29, 2020 at 3:00 AM <Yun> ***@***.***> wrote: Hi, Thanks for your code! Recently I need to re-train this network, on another dataset, HKU-IS, the model converges good, but I don't know when to stop it... the loss now in your code is 0.099 and 2.83 respecitively... can you tell us when to stop the training when you train the BASNet using DUT-TR dataset? — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#46>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ADSGORIOMALB2TX4CYTRL73SIGO2JANCNFSM4R5QOWIA> .

-- Xuebin Qin PhD Department of Computing Science University of Alberta, Edmonton, AB, Canada Homepage:https://webdocs.cs.ualberta.ca/~xuebin/

[BarCodeReader]

yeah I know how to implement a eval dataloader and it would be fine if we have it.
i think i also don't have enough GPU RAM to do so, just like you mentioned in the paper, hahaha.

yeah, i also now plot the log(loss) as you mentioned.

Thanks for the code again!